1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
|
//! Linux process code for getting process data via `/proc/`.
//! Based on the [procfs](https://github.com/eminence/procfs) crate.
use std::{
fs::File,
io::{self, BufRead, BufReader, Read},
path::PathBuf,
};
use anyhow::anyhow;
use libc::uid_t;
use once_cell::sync::Lazy;
use rustix::{
fd::OwnedFd,
fs::{Mode, OFlags},
path::Arg,
};
use crate::Pid;
static PAGESIZE: Lazy<u64> = Lazy::new(|| rustix::param::page_size() as u64);
#[inline]
fn next_part<'a>(iter: &mut impl Iterator<Item = &'a str>) -> Result<&'a str, io::Error> {
iter.next()
.ok_or_else(|| io::Error::from(io::ErrorKind::InvalidData))
}
/// A wrapper around the data in `/proc/<PID>/stat`. For documentation, see [here](https://man7.org/linux/man-pages/man5/proc.5.html).
///
/// Note this does not necessarily get all fields, only the ones we use in bottom.
pub(crate) struct Stat {
/// The filename of the executable without parentheses.
pub comm: String,
/// The current process state, represented by a char.
pub state: char,
/// The parent process PID.
pub ppid: Pid,
/// The amount of time this process has been scheduled in user mode in clock ticks.
pub utime: u64,
/// The amount of time this process has been scheduled in kernel mode in clock ticks.
pub stime: u64,
/// The resident set size, or the number of pages the process has in real memory.
pub rss: u64,
}
impl Stat {
#[inline]
fn from_file(mut f: File, buffer: &mut String) -> anyhow::Result<Stat> {
// Since this is just one line, we can read it all at once. However, since it might have non-utf8 characters,
// we can't just use read_to_string.
f.read_to_end(unsafe { buffer.as_mut_vec() })?;
let line = buffer.to_string_lossy();
let line = line.trim();
let (comm, rest) = {
let start_paren = line
.find('(')
.ok_or_else(|| anyhow!("start paren missing"))?;
let end_paren = line.find(')').ok_or_else(|| anyhow!("end paren missing"))?;
(
line[start_paren + 1..end_paren].to_string(),
&line[end_paren + 2..],
)
};
let mut rest = rest.split(' ');
let state = next_part(&mut rest)?
.chars()
.next()
.ok_or_else(|| anyhow!("missing state"))?;
let ppid: Pid = next_part(&mut rest)?.parse()?;
// Skip 9 fields until utime (pgrp, session, tty_nr, tpgid, flags, minflt, cminflt, majflt, cmajflt).
let mut rest = rest.skip(9);
let utime: u64 = next_part(&mut rest)?.parse()?;
let stime: u64 = next_part(&mut rest)?.parse()?;
// Skip 8 fields until rss (cutime, cstime, priority, nice, num_threads, itrealvalue, starttime, vsize).
let mut rest = rest.skip(8);
let rss: u64 = next_part(&mut rest)?.parse()?;
Ok(Stat {
comm,
state,
ppid,
utime,
stime,
rss,
})
}
/// Returns the Resident Set Size in bytes.
#[inline]
pub fn rss_bytes(&self) -> u64 {
self.rss * *PAGESIZE
}
}
/// A wrapper around the data in `/proc/<PID>/io`.
///
/// Note this does not necessarily get all fields, only the ones we use in bottom.
pub(crate) struct Io {
pub read_bytes: u64,
pub write_bytes: u64,
}
impl Io {
#[inline]
fn from_file(f: File, buffer: &mut String) -> anyhow::Result<Io> {
const NUM_FIELDS: u16 = 0; // Make sure to update this if you want more fields!
enum Fields {
ReadBytes,
WriteBytes,
}
let mut read_fields = 0;
let mut reader = BufReader::new(f);
let mut read_bytes = 0;
let mut write_bytes = 0;
// This saves us from doing a string allocation on each iteration compared to `lines()`.
while let Ok(bytes) = reader.read_line(buffer) {
if bytes > 0 {
if buffer.is_empty() {
// Empty, no need to clear.
continue;
}
let mut parts = buffer.split_whitespace();
if let Some(field) = parts.next() {
let curr_field = match field {
"read_bytes:" => Fields::ReadBytes,
"write_bytes:" => Fields::WriteBytes,
_ => {
buffer.clear();
continue;
}
};
if let Some(value) = parts.next() {
let value = value.parse::<u64>()?;
match curr_field {
Fields::ReadBytes => {
read_bytes = value;
read_fields += 1;
}
Fields::WriteBytes => {
write_bytes = value;
read_fields += 1;
}
}
}
}
// Quick short circuit if we have already read all the required fields.
if read_fields == NUM_FIELDS {
break;
}
buffer.clear();
} else {
break;
}
}
Ok(Io {
read_bytes,
write_bytes,
})
}
}
/// A wrapper around a Linux process operations in `/proc/<PID>`.
///
/// Core documentation based on [proc's manpages](https://man7.org/linux/man-pages/man5/proc.5.html).
pub(crate) struct Process {
pub pid: Pid,
pub uid: Option<uid_t>,
pub stat: Stat,
pub io: anyhow::Result<Io>,
pub cmdline: anyhow::Result<Vec<String>>,
}
#[inline]
fn reset(root: &mut PathBuf, buffer: &mut String) {
root.pop();
buffer.clear();
}
impl Process {
/// Creates a new [`Process`] given a `/proc/<PID>` path. This may fail if the process
/// no longer exists or there are permissions issues.
///
/// Note that this pre-allocates fields on **creation**! As such, some data might end
/// up "outdated" depending on when you call some of the methods. Therefore, this struct
/// is only useful for either fields that are unlikely to change, or are short-lived and
/// will be discarded quickly.
pub(crate) fn from_path(pid_path: PathBuf) -> anyhow::Result<Process> {
// TODO: Pass in a buffer vec/string to share?
let fd = rustix::fs::openat(
rustix::fs::CWD,
&pid_path,
OFlags::PATH | OFlags::DIRECTORY | OFlags::CLOEXEC,
Mode::empty(),
)?;
let pid = pid_path
.as_path()
.components()
.last()
.and_then(|s| s.to_string_lossy().parse::<Pid>().ok())
.or_else(|| {
rustix::fs::readlinkat(rustix::fs::CWD, &pid_path, vec![])
.ok()
.and_then(|s| s.to_string_lossy().parse::<Pid>().ok())
})
.ok_or_else(|| anyhow!("PID for {pid_path:?} was not found"))?;
let uid = {
let metadata = rustix::fs::fstat(&fd);
match metadata {
Ok(md) => Some(md.st_uid),
Err(_) => None,
}
};
let mut root = pid_path;
let mut buffer = String::new();
// NB: Whenever you add a new stat, make sure to pop the root and clear the buffer!
let stat =
open_at(&mut root, "stat", &fd).and_then(|file| Stat::from_file(file, &mut buffer))?;
reset(&mut root, &mut buffer);
let cmdline = cmdline(&mut root, &fd, &mut buffer);
reset(&mut root, &mut buffer);
let io = open_at(&mut root, "io", &fd).and_then(|file| Io::from_file(file, &mut buffer));
Ok(Process {
pid,
uid,
stat,
io,
cmdline,
})
}
}
#[inline]
fn cmdline(root: &mut PathBuf, fd: &OwnedFd, buffer: &mut String) -> anyhow::Result<Vec<String>> {
open_at(root, "cmdline", fd)
.map(|mut file| file.read_to_string(buffer))
.map(|_| {
buffer
.split('\0')
.filter_map(|s| {
if !s.is_empty() {
Some(s.to_string())
} else {
None
}
})
.collect::<Vec<_>>()
})
.map_err(Into::into)
}
/// Opens a path. Note that this function takes in a mutable root - this will mutate it to avoid allocations. You
/// probably will want to pop the most recent child after if you need to use the buffer again.
#[inline]
fn open_at(root: &mut PathBuf, child: &str, fd: &OwnedFd) -> anyhow::Result<File> {
root.push(child);
let new_fd = rustix::fs::openat(fd, &*root, OFlags::RDONLY | OFlags::CLOEXEC, Mode::empty())?;
Ok(File::from(new_fd))
}
|